Explanation Versus Meta-Explanation: What Makes a Case More Convincing

Comparative analysis of the roles of explanation and meta explanation is conducted assessing the validity of explanation exchanged between human agents. Meta explanation links the overall structure of a current scenario with that of previously learned scenarios of multi agent interaction. The scenario structure includes communicative actions of involved agents and argumentation attack relations between the subjects of these actions. Object level explanation is based on a traditional machinery to handle argumentative structure of a dialogue, assessing the plausibility of individual claims. To assess plausibility of customer complaints, we relate them to the classes of valid (consistent, genuine) and invalid (inconsistent, include attempts to get compensation from a company, or expressing a bad mood). Evaluation of contribution of each explanation level shows that both levels of explanation are essential for assessment of whether a multi agent scenario as described by an agent is plausible or not. We demonstrate that meta explanation in the form of machine learning of scenario structure should be augmented by conventional explanation by based arguments for individual claims. We also de ne a ratio between object level and meta explanation as relative accuracy of plausibility assessment based on former and latter sources. We then observe that groups of scenarios can be characterized based on a speci c ratio between object level and meta level explanations in a phase space; such ratio is an important parameter of human behavior associated with explaining in a dialogue.